cyberneticlibrary

Score skills for autonomous loop optimization

eval-autoresearch-fitskillsetup L23
richfrem/agent-plugins-skills
What it does

Score skill viability for autonomous Karpathy-loop optimization

Best for

Identifying which skills are ready for autonomous improvement loops by assessing eval robustness

Inputs
  • · skill description
  • · current evals
Outputs
  • · autoresearch fit score with red/green/refactor cycle recommendations
Requires
  • · os-skill-improvement
  • · eval-gate
Preconditions

Skill has evals.json; skill purpose is deterministic/measurable

Failure modes
  • · skill too vague (improve description)
  • · eval cases too few (add more)
Trust signals
  • · Karpathy loop compatibility check
  • · eval-gate integration
  • · RED-GREEN-REFACTOR cycle